This paper presents a divide-and-conquer method that finds a near-optimal distribution of sensing locations in a very efficient manner for the free-form surface digitization process. We formulate the sensing localization issue as a next-best-point problem. We transform the uncertainty of a reconstructed B-spline surface into a higher-dimensional uncertainty surface. This further allows the use of convex hull and subdivision properties of B-spline surfaces in the NBP based sensing localization algorithm. It thus dramatically reduces the search time for determining the next best sensing location. Experimental examples demonstrate that the algorithm compares favorably to existing algorithms and, due to its high efficiency, supports both off-line and on-line sensing planning.

This content is only available via PDF.
You do not currently have access to this content.